730 



THE AMERICAN NATURALIST [Vol. L 



It is to be observed that a cross makes a profound dif- 

 ference in the effectiveness of return selection. Crossing 

 has so modified the germ plasm that rats from the minus 

 race immediately, without any gradual return to the "0" 

 grade, repeated the history of the plus race. Further, 

 plus selection was carried on in this new race. Castle 

 (Castle and Phillips, :14, p. 21) emphasizes the fact that 

 this race is free from the objection urged against the 

 main experiment, namely that the closest inbreeding was 

 not carried out. Further interest in this closely inbred 

 race lies in the fact that, although it starts out with a 

 curve almost identical with the first generations of the 

 plus race, the rate of advance falls off faster than it does 

 in the main plus race. One may suppose that the cross 

 produced an F 2 in which some rats had a degree of het- 

 erozygosity similar to that which existed in the original 

 unselected stock; a closer inbreeding reduced the hetero- 

 zygosity more rapidly. 



12. The earlier generations of the plus race when 

 crossed with wild are only slightly reduced in pigmenta- 

 tion. In Table 43, Castle and Phillips ( :14, p. 48) show, 

 among other things, the averages of hooded grandchil- 

 dren extracted from a cross with Irish. In comparison 

 with these are placed averages specified to be of offspring 

 from the same grade parents and the same generation of 

 the uncrossed selected race. References to the proper 

 tables of the uncrossed selected races show practical 

 agreement with the averages as quoted in this Table. 

 In Table 42, which gives corresponding results of crosses 

 with wild, three of the averages of the uncrossed races 

 are taken from the same generation as the parents 

 crossed, and three seem to be taken from the follow- 

 ing generation. It is a matter of importance to have 

 correct standards for judging the modifications due to 

 crossing. There might be a question whether one should 

 use the average of the generation from which the 

 parents came, or the following one; but in either case 

 the use should be constant. Although the averages of 



